Some Consequences of a Deletion Analysis of Null Subjects Ian Roberts Downing College, University of Cambridge

Transcription

1 Some Consequences of a Deletion Analysis of Null Subjects Ian Roberts Downing College, University of Cambridge igr20@cam.ac.uk 1. Holmberg (2002) (1) Rizzi (1986a): a. pro must be licensed b. pro must be identified (2) IP pro i I I i [3pl] (3) If Agr is inherently unspecified pro can t be identified by Agr. If pro is inherently unspecified, Agr can t be specified by pro. he GB theory of pro and Chomsky s 1995 theory of features can t both be right. (4) Hypothesis A: in null-subject languages, Agr is interpretable. Spec is therefore either absent or filled by an expletive. his expletive is usually silent (i.e. pro), but wouldn t have to be: possibility of overt expletive subject. Hypothesis B: a silent pronoun (pro) with interpretable features occupies Spec and functions just like an overt pronoun. his implies that there is no possibility of filling Spec with an overt expletive: no possibility of over expletive subject. (5) Finnish has null subjects and an overt expletive pronoun, sitä: a. Puhun englantia. speak-1sg English I speak English b. Sitä meni hyt hullusti. EXPL went now wrong Now things went wrong. (6) Expletive does not cooccur with referential null subjects: a. *Sitä puhun englantia. EXPL speak-1sg English b. Oletteko (*sitä) käyneet Pariisissa? have-2pl-q EXPL visited Paris? herefore Hypothesis B is right: pro occupies Spec. But can t be licensed/identified as in Rizzi (1986a) (see (3)). Since this element is like an overt pronoun in all respects except phonological realisation, assume it is a deleted (or PF-non-realised) pronoun: the deletion analysis of null subjects. 1

2 2. A variant (7) canto ( I sing ): [id, i1, isg] vp [ud, u1, usg, EPP] (D) ([id, i1, isg]) ((Cat) = trace/copy of Cat). (NB features may not directly belong to but V, and be carried there thanks to s V- attracting property). (8) a. Agree: α Agrees with β iff α c-commands β and α and β are non-distinct in features (and there is no category γ equivalent to α which c-commands β but not α). b. Where α Agrees with β, α is the Probe and β is the Goal. In (7), probes D, and so the pronoun s interpretable features check the corresponding uninterpretable features of. s EPP feature triggers movement of D(P) to Spec from its vp-internal merged position. he trace/copy of D is deleted at PF. (9) Deletion under feature-identity: α deletes under identity of features with β only if β Agrees with α. (NB this differs from VP-ellipsis, which doesn t require either Agree or feature-identity, but a weaker structural condition and non-distinctness). So both copies of D are able to delete in (7). D-deletion takes place after deletion of s EPP feature by movement. After deletion of s EPP feature by movement, and deletion of D s Nom-feature under Agree, this gives rise to null subjects by D- deletion. (10) a. Uninterpretable = unvalued and Agree = feature-valuing (Chomsky (2001)). b. Features are attribute-value pairs, e.g. [pers: 3], and uninterpretable features lack a value, e.g. [pers: ]. Unvalued features must be valued before deletion under identity can take place. (11) he null-subject parameter: is +/- uninterpretable D (pace Holmberg). 2

3 (12) In non-null-subject languages, lacks a D-feature, and so the relevant featureidentity does not obtain after D-movement. French: [id, i1, isg] [u1, usg, EPP] Here D is realised as je (can t be deleted due to lack of identity with ). (13) Rich agreement: has a D feature just where it has a complete set of person-number features; in languages where person-number features are only partially specified, may have such features, but no D feature. Partial specification is a concrete PF matter. [(14) It appears that a complete set of person-number features actually tolerates one syncretism, including zero-endings (Roberts (1993b:127f.): Early OF Rumanian Icelandic Modern French chant cînt tel chante chantes cînţi telur chantes chante(t) cîntă telur chante chantons cîntăm teljum chantons chantez cîntaţi teljið chantez chantent cîntă telja chantent 6 distinct 5 distinct 5 distinct 3 or 4 distinct] (15) he East Asian problem: All languages which allow discourse pro-drop allow (robust) bare NP arguments ( Discourse Pro-Drop Generalization, omioka (2003:336)). Null pronouns in Discourse Pro-Drop languages are simply the result of N - Deletion/NP-Ellipsis without determiner stranding (ibid). 3. Null subject pronouns and clitic pronouns 3.1 Non-subject clitics (16) Case: all Case features are uninterpretable; the subject s Nominative feature is checked by the Agree relation with s phi- features, but Acc and Dat are not (Chomsky (2001)). Clitics have Acc (le) or Dat (lui), distinct in feature-content from and unable to delete (even though they appear to be attracted to ; Belletti (1999), but cf. Ledgeway (2003)). 3

4 3.2 Subject clitics in Northern Italian dialects (17) A doubling variety: Fiorentino (Brandi & Cordin (1989)): (E) parlo u parli E parla La parla Si parla. Vu parlate. E parlano. Le parlano. -- both clitics and verb-endings suffice to delete the subject under identity by (14). (18) Syncretism of subject clitics: San Michele a. (Friulian; Poletto (2000)): I mangi I ti mangis A l mangia I mangin I mangè A mangin -- verbal inflection is rich by the definition in (14) (one syncretism). (19) A complementary system: Paduan (Poletto (1993)): vegno te vien el vien vegnemo vegni i vien (20) [id, i1, ipl] [ud, u1, upl, EPP] NB has ud here, but not in French, so deletion under identity can take place. o make this consistent with (14), the subject clitics must count as -features (pace Cardinaletti & Repetti (2003), but in line with Rizzi (1986b), Brandi & Cordin (1989), Poletto (2000) etc.). (21) What would French look like if it had [+D]? Either fully null subject, or, if we take WYSIWYG approach to inflectional morphology, then perhaps French only has [Pers:1, Pers:2, Num:Pl], realising the inflections ons and ez. he other values (Num: Sg, Pers: 3) may be filled in at LF by default. 4

5 (22) Early 17th-century French (Maupas (1607), cited in Brunot (1905, III:477), Roberts (1993:215)): a. J ay receu les lettres que m avez envoyees I ve received the letters that me-have.2pl sent I ve received the letters you sent me. b. Vous voyez qu avons soin de vous. You see that have.1pl care of you You see that we care about you. (But this is not consistent with (13, 14)). 4. Proclitic and enclitic paradigms of subject clitics (23) If interrogative sentences are formed via subject inversion, (i) the number of enclitic pronouns found in interrogative sentences is equal to or greater than the number of proclitic pronouns in declarative sentences, and (ii) the subject pronouns found in proclitic position are also found in enclitic position. (Renzi & Vanelli s (1983) Generalisation 9; see also Poletto (2000:42ff.)) (24) Paduan: Proclitics: --, te, el/la, --, --, i/le (see (19)) Enclitics: i, to, lo/la, i, o, li/le (25) Inversion as -to-c movement (see Poletto (2000:42ff.)): CP C [iq, EPP ] [ipers, inum, id] [uq, upers, unum, ud] and D are not identical in feature content, and null subjects are thus impossible in the inversion context even where all the other conditions for null subjects are met. (26) Montesover (rentino) (Poletto (2000:54)): a. Piove. rains It s raining. b. Piove-l? rains-scl Is it raining? 5

6 5. Null subjects and V2 environments 5.1 Rhaeto-Romance (Haiman (1988), Haiman & Benincà (1992:167ff.), Poletto (2000, Chapter 3)) With the exception of Ladin.., all Rhaeto-Romance dialects distinguish three persons in both the singular and the plural [in the present indicative IGR] (Haiman (1988:360)). So all varieties should allow null subjects freely by (14), but they do not (Haiman (1988:384)). (27) Surselvan is V2+full postverbal subject pronouns and optional expletive deletion and 2sg/2pl deletion in Spec (like OF for [Pers:2]): i. Ed aschia fa el il patg cul nausch and so makes he pact with-the devil And so, he makes a pact with the devil. ii. iii. Avon casa ei-s-i mats. before house has-s-it boys Before the house, some boys are standing. em emprema lingia ei vegniu examinau il stan tecnic dils vehichels in first line is come examine the state technical ofthe vehicles First, the technical condition of the vehicles was examined. (28) he other Swiss varieties (Puter, Vallader, Sutselvan and Surmeiran) are V2 + enclitic postverbal subject pronouns which delete according to phonological rules, e.g. the general ban on antepenultimate stress, which causes vocalic enclitic pronouns to delete where the preceding verb has penultimate stress. (29) Badiot and Gardena phonological rules delete both pro- and enclitic subjects in V2 contexts:) i. V deletes before V-initial verb (Badiot): ùn tut pert a chësc concurs have (1pl) taken part to this competition We have taken part in this competition. ii. iii. Pronouns drop before all aux: (i) sun sta dër cuntenĉ (we) are been very glad we have been very glad he three-syllable rule (no antepenultimate stress on verb): ci podésson-ø pa fa por os? what could-1pl then do for you? What could we do for you, then? 6

7 (30) Other varieties (Ladin, Friulian) aren t V2 and allow null subjects (Haiman (1988:385)). (31) What prevents deletion of subject pronoun under identity with in these varieties? A crucial feature must belong to C not to : CP C [upers/num, EPP] [ipers, inum, id] [upers/num, ud, EPP] (NB explains why Icelandic is not a null-subject language see (14)). 5.2 Old French (32) Null subjects allowed just where we have V2 (Vanelli, Renzi, Benincà (1986), Adams (1987), Vance (1988, 1997), Roberts (1993)): a. Si en orent -- moult grant merveille hus of-it had-3pl very great marvel (Merlin, 1) So they wondered very greatly at it. (Roberts (1993:124f.)) b. Ainsi s acorderent -- que il prendront par nuit. hus they-agreed that they will-take by night his they agreed that they would take by night (Le Roman du Graal, B. Cerquiglini (ed), Union Générale d Editions, Paris, 1981, 26; Adams (1987b:1), Roberts (1993:84)) -- whatever triggers V2 must also license deletion under identity in OF but not in RR. Agreement features? (33) OF main clause at end of phase: CP C+ [EPP, upers, unum, ud] [ipers, inum, id] () Complementisers lack these features and hence block deletion under identity. 7

8 Could V2/null-subject systems arise through generalisation of phonological deletion rules (i.e. dropping of conditions from rules such that pure deletion under identity is all that remains)? 6. Conclusion A deletion analysis of null subjects has some merit (general relation to clitics, account of proclisis/enclisis differences, rich-agreement/morphological uniformity) and may be forced by the mechanisms of the current version of the theory. Questions regarding some aspects of interaction with V2 remain, although there seem to be two distinct types of interaction. 8